Background
Inaccessibility and inaccuracies limit histopathologic diagnosis of Kaposi sarcoma (KS) in Africa. We hypothesized that quantification of KSHV DNA from suspicious skin lesions could distinguish KS from non-KS lesions, thus eventually enabling creation of a diagnostic test that would limit need for histopathologic interpretation. In our earlier training population, quantification of lesional KSHV DNA by qPCR (using sub-optimally transportable Ct values) had high sensitivity and specificity for the diagnosis of KS. We now assess diagnostic accuracy of qPCR in an external validation population and using a more transportable KSHV copy number readout.
Methods
Using identical processes for both the training (N=506 participants) and validation populations, we performed skin punch biopsies among consecutive patients referred for suspected KS in Uganda. One-half of the biopsy had histopathology performed, and the other half was stored in RNAlater and later underwent DNA purification in the U.S. and testing for KSHV ORF 26 DNA in two replicates by qPCR. With histopathology as gold standard and ROC curves, we assessed the performance characteristics of the mean of the replicates for the diagnosis of KS. An optimal KSHV DNA copy number cutoff that was earlier derived in the training population was assessed in the validation population.
Results
The validation population consisted of 427 participants with skin biopsies; 34% were women, and 91% were HIV-infected. Histopathology revealed that 338 biopsies were KS and 89 were not. Using the previously-derived threshold of 161 copies of KSHV DNA per 5 µl input into the reaction, which in our training population yielded 91% sensitivity and 91% specificity for KS diagnosis (accuracy=91% and area under the curve (AUC) = 0.96), qPCR in the validation population yielded 93% sensitivity, 93% specificity, 93% accuracy and an AUC of 0.93 (Figure).
Conclusions
In an external validation population in Uganda, qPCR of lesional KSHV DNA had high sensitivity and specificity for diagnosis of KS, performing similarly to our training population. Because of ambient endemic KSHV infection, mere qualitative detection of KSHV DNA in skin lesions is non-specific for KS diagnosis. Instead, precise quantification of KSHV DNA is needed, and a KSHV DNA copy number threshold that balances sensitivity and specificity is now established. This validation of qPCR performance motivates a point-of-care nucleic acid-based diagnostic test for KS, which we are currently developing.